Qwen3-4B-Base
A base language model in Alibaba's Qwen series with 4B parameters, suitable for research and fine-tuning.
Timeline (last 90 days)
Reproduction experiments on Qwen3-4B-Base show that MCPO also improves over GRPO, with all three seeds positive and an average improvement of 4.75 points.
In the RLVR experiments on Qwen3-4B-Base, with both the SGD+EMA and AdamW optimizers, configurations choosing a larger momentum coefficient showed an advantage.
Fiorillo v0.5 uses Qwen3-4B-Base as its backbone, with a low-rank adapter (LoRA) and a dedicated decision head stacked on top, fine-tuned exclusively for the EI task.
All Facts (3)
Reproduction experiments on Qwen3-4B-Base show that MCPO also improves over GRPO, with all three seeds positive and an average improvement of 4.75 points.
50%UnverifiedIn the RLVR experiments on Qwen3-4B-Base, with both the SGD+EMA and AdamW optimizers, configurations choosing a larger momentum coefficient showed an advantage.
50%UnverifiedFiorillo v0.5 uses Qwen3-4B-Base as its backbone, with a low-rank adapter (LoRA) and a dedicated decision head stacked on top, fine-tuned exclusively for the EI task.
50%